2.2. EL JUEGO
2.2.3. CARACTERÍSTICAS DEL JUEGO
Introduction
The visual system takes time to process and store incoming visual information. Attention thus acts as a filter to select the most important, task critical information at the expense of subsequent
information. Perhaps the most well-known example of this is the attentional blink (AB)
phenomenon; when two visual target items are embedded as part of a rapid series of items, the second target will often be missed if it occurs 200-500 ms after the first target (Raymond, Shapiro, &
Arnell, 1992). When the first target (T1) is task irrelevant, the second is generally reported, demonstrating that the loss of the second target (T2) is due to the attention devoted to the first, rather than a pure sensory deficient. Thousands of articles have been written examining the nature of the attentional blink deficient over the past 20 years (See Dux & Marois, 2009; Martens & Wyble, 2010), though the nature of the AB remains somewhat unresolved. In a recent review, Hanslmayr and colleagues put forward the proposition that the attentional blink is, in part, due to a shift from an externally oriented state to an internally oriented state of visual processing (Hanslmayr et al., 2011). An externally oriented state is necessary for the collection of new incoming information with high fidelity, whereas an internally oriented state is ideally suited for maintaining selected objects in working memory and matching to held template information. The authors propose that alpha activity forms part of the substrate of such as process. During an internal state, such as triggered following the first target, increased occipital alpha inhibits the processing of further sensory information through pulsed inhibition, while promoting cross-regional coupling of retained information. Endogenous alpha phase had already been demonstrated to partially predict the success or failure of masked single target perception (Hanslmayr et al., 2007; Mathewson et al., 2009). This framework in relation to the AB was supported by studies which report greater resting state alpha power in individuals with larger AB magnitudes (MacLean, Arnell, & Cote, 2012) as well
28 as greater alpha phase synchronization just before the second target when the second target is missed (Kranczioch, Debener, Maye, & Engel, 2007; Zauner et al., 2012).
Hanslmayr and colleagues further point out that the lion’s share of previous work on the attentional blink has been conducted using an RSVP rate of ~8-12 Hz. As rapid alpha-band visual presentation is known to result in the phase locking of matching frequencies in the occipital lobe, it is possible that the RSVP stream entrains alpha activity to a phase which is poor for perception. This theory predicts that any disruption of such alpha “entrainment” should facilitate processing of the second target.
Indeed, Martin and colleagues reported that the introduction of temporal jitter, by varying the ISI of an RSVP stream (17 to 153 ms ISI, 12 Hz average rate) before T1 and T2 substantially reduces the magnitude of the blink (Martin et al., 2011), which could be attributed the disruption of alpha entrainment. No difference in AB magnitude between regular and irregular stimulation was found in a subsequent study of the AB (Zauner et al., 2012), though the authors argued that the irregular stimulation used was insufficient to disrupt alpha entrainment (+-20 ms). Any explanation for the attentional blink must also take into account the “skeletal blink”; the presence of the attentional blink when only four items are presented (two targets and two masks) (Duncan, Ward, & Shapiro, 1994; McLaughlin, Shore, & Klein, 2001). The presence of the skeletal blink, however, could be accounted by the presence of endogenous alpha activity in the absence of rapid stimulation, leading to an endogenous “internal” state at each target onset, and the loss of T2 processing.
With this framework in mind, Elwyn Martin, Simon Hanslmayr, Jim Enns, Alejandro Lleras and Kimron Shapiro proposed that only an alpha-band (8-12 Hz) RSVP stream will lead to a substantial attentional blink (unpublished). To test this hypothesis, Martin and colleagues presented an RSVP stream of task irrelevant grey letters at theta, alpha, beta and gamma rates. Two pop-out red targets were embedded in-phase with the stream with lags of ~100, 300 or 700 ms, adjusted to maintain the rhythmicity of the stream until the occurrence of T2. After T2, the RSVP in all conditions reverted to a rate of 10 Hz. Initial tests appeared to show the presence of an attentional blink only for the alpha
29 (10 Hz) stimulation condition. Unfortunately, this result may have resulted from a subsequently discovered error in stimulus presentation.
Experiment 1
The design and methods used in Experiment 1 resulted as a follow-up to unpublished work previously conducted by Elwyn Martin, Simon Hanslmayr, Jim Enns, Alejandro Lleras, and Kimron Shapiro. Experiment 1 is a redesign of this work with moderate modifications. 1
In Experiment 1a (N = 24), participants had to identify two target red letters that appeared in rapid succession embedded in an RSVP stream of distractor black letters (Figure 2.1). The two red target letters (T1 and T2) had 3 possible SOAs of ~100 ms, ~300 ms , or ~700 ms, with the middle SOA selected to match the strongest time point of the attentional blink (lowest T2 performance). Black distractor letters were presented at 4 different speeds (6.3 Hz, 10 Hz, 16 Hz, and 36 Hz) separated in a block design. Experiment 1b (N = 24) was the same as Experiment 1a, except participants were told to ignore the first red letter. Experiment 1b served as a control experiment to demonstrate the attentional nature of the blink observed in Experiment 1a, with the expectation of no blink when the first red item is not a target.
Methods
1 The work of the current chapter regarding the attentional blink was conducted in collaboration with Elwyn Martin, Simon Hanslmayr, Jim Enns, Alejandro Lleras and Kimron Shapiro. I was asked to replicate the results of the previous experiment with a new presentation script, faster LCD monitor, and modifications on the previous design. Specifically, to address a concern that the absence of a non-alpha blink may have been related to the temporal irregularity of a frequency switch after T2 in all but the alpha condition, I changed the design such that the post-T2 period maintained the frequency of the preceding stream. All other modifications were minor. The hypothesis remained the same; only alpha-band stimulation would lead to a substantial attentional blink. The design of Experiment 2 and the final interpretation of the results are my own.
30 Twenty-four participants (mean age: 19.5 years, 21 female) took part in Experiment 1a. Participants were to be excluded if overall performance across all conditions fell below 50% (Chance: 33%), and by this criteria no participants were excluded from analysis.
Stimuli were presented on a 27” ASUS VG278HE LCD monitor with a grey-to-grey response time of 2 ms, set to a refresh rate of 144 Hz. Participants were seated approximately 70 cm from the display.
All letter stimuli were presented in Arial Bold 36 point font (~ 1° visual angle in diameter) for three frames (21 ms) each against a grey background (RGB: [127 127 127]), all distractor letters were black (RGB: [0 0 0]) and all target letters were red (RGB:[255 0 0]). The first target (T1) was always one of three letters (B, G or S); the second target (T2) was one of a different set of three letters (X, K, or Y).
Each trial consisted of the central, serial presentation of distractor letters (the remaining 22 non-target letters of the English alphabet, randomly selected with replacement) and two non-target letters, presented at one of four different frequencies (6.3 Hz, 10 Hz, 16 Hz, or 36 Hz, blocked). Within a trial, the time between each letter presentation was held constant, such that the letter series was
completely isochronous. A trial began with 500 ms of a blank grey screen, followed by the serial presentation of distractor stimuli for ~1000 ms before the presentation of T1, and for ~550 ms after the presentation of T2. The time interval between the first and second target (lag) was manipulated to be approximately 100, 300 or 700 ms for each sequence frequency. The exact time intervals between T1 and T2 for each sequence frequency and lag can be found in Table 1, and the
corresponding number of intervening items can be found in Table 2. A black asterisk immediately following the final distractor item cued the participant to report T1 and T2, and the next trial followed immediately after T2 report. A diagram of the trial timing can be found in Figure 2.1.
31
Figure 2.1 Trial design of Experiment 1. Each vertical line represents the timing of presentation of a single letter. Letters were presented at one of four different frequencies on a given block, with block order counterbalanced across
participants. In Experiment 1a, T1 was to be reported, in Experiment 2, participants were instructed not to report T1.
Table 1: Exact interval parameters for each frequency (ms) Lag
Frequency 100ms 300ms 700ms
36.0 Hz (Gamma) 111 306 694
16.0 Hz (Beta) 125 313 688
10.3 Hz (Alpha) 97 292 681
6.3 Hz (Theta) 160 319 639
Table 2: Number of stimuli between T1 and T2, for each frequency x lag condition.
Lag
Frequency 100ms 300ms 700ms
32
36.0 Hz (Gamma) 3 10 24
16.0 Hz (Beta) 1 4 10
10.3 Hz (Alpha) 0 2 6
6.3 Hz (Theta) 0 1 3
Participants were explicitly told and shown the possible letter identities of T1 and T2 before the experiment, and were asked to identify and report the two red letter targets in order. Only
responses corresponding to one of the three possible letters for each target position were accepted, thus participants had to select from three independent, alternative choices for each target position.
Thus, chance accuracy for each target was 33%, and order reversals were not possible. Participants selected each letter by pressing the corresponding button on the keyboard.
Each participant completed four blocks of 81 trials, one block for each sequence frequency, for a total of 324 trials. Each block contained a counterbalanced and randomly ordered set of 81 trials from a 3x3x3 design (T1 identity, T2 identity, Lag). Frequency block order was fully counterbalanced across participants. Participants were given a brief self-paced break between blocks.
The methods and procedures used in Experiment 1b were identical to Experiment 1a, except that the participants was instructed to only report the second red letter target, and only a single
response corresponding to one of the possible T2 identities was recorded. Twenty-four participants (mean age: 19.1 years, 22 female) participated in Experiment 1b.
Analysis
To examine differences in the magnitude of the attentional blink across frequency in Experiment 1a, we conducted a repeated measures one-way ANOVA of AB magnitude, defined as the % correct difference in performance between T2 accuracy for all trials in which T1 was correctly identified (T2|T1) at Lag 700 minus Lag 300. T2|T1 accuracies at Lag 100 are included to examine the extent of
33 Lag 1 sparing in the current paradigm, though not included in the calculation of AB magnitude. We also conducted a repeated measures one-way ANOVA on T1 accuracy, regardless of lag, to examine the effect of frequency on target visibility. For Experiment 1b, a repeated measures one-way ANOVA of general target accuracy (second red letter) was conducted. To determine the change in AB
magnitude resulting from ignoring T1, a mixed effects two-way ANOVA (Frequency x Group) was conducted comparing the magnitude of the blink between Experiment 1a and 1b. All reported statistics were Greenhouse-Geisser corrected for violations of sphericity and all pairwise statistics were Bonferroni corrected for multiple comparisons. We also performed independent one-sample t-tests on each frequency bin to determine if the AB magnitude was significantly larger than chance for each frequency condition.
Results
In Experiment 1a, we found a significant main effect of Frequency (F(3,69) = 7.13, p = .002, partial η2
= .643) on AB magnitude (See Figures 2.2 and 2.3). Pairwise comparisons revealed a significantly larger blink for the 10.3 Hz and 16 Hz conditions than the 6.3 Hz and 36 Hz conditions (all p < .05), with no other significant differences between conditions. AB magnitude was significantly larger than zero at 10.3 Hz (p = .012) and 16 Hz (p = .006), while AB magnitude was not significantly different from zero at 6.3 Hz (p = 0.56) and 36 Hz (p = .22). Thus, an AB was observed at both 10 Hz and 16 Hz, contrary to the initial alpha entrainment hypothesis. T1 accuracy was significantly different between conditions (F(3,69) = 43.2, p < .001, partial η2 = .831), with accuracy falling monotonically as a function of frequency, as would be expected due to an increase in forward and backward masking of the task-irrelevant stimuli with increasing frequency.
34
Figure 2.2 The results of Experiment 1a. Error bars represent the across participant standard error.
In Experiment 1b, when participants were instructed not to report T1, the overall magnitude of the blink was significantly reduced compared to Experiment 1a (F(1,46) = 12.6, p = .001, partial η2 = .215) (See Figure 2.3), demonstrating that the blink produced in the current experiment was due in large part to the allocation of attention and the top-down selection of T1, rather than pure bottom-up stimulus properties, consistent with previous AB literature.
35
Figure 2.3 AB magnitudes from Experiment 1a and 1b and T1 accuracy from Experiment 1a. Error bars represent the across participant standard error.
Thus, the results were inconsistent with the hypothesis that only stimulation at alpha band (10 Hz) would produce an attentional blink, as an equal or greater blink was occurred when items were presented at a rate of beta band (16 Hz). However, the results of Experiment 1 do not preclude a substantial role of entrainment in producing the attentional blink. For instance, it is possible that both stimulations at 10 Hz and 16 Hz lead to inhibitory entrainment at 10 Hz and 16 Hz, respectively, or that both lead entrainment of a single critical band of activity between 10 and 16 Hz (i.e. .,high-alpha, low-beta). The synchronization of both alpha and beta-band activity are often altered by sensory events in concert (Klimesch, 2012) and both bands are reportedly correlated with T2 performance on a trial-by-trial basis (Glennon, Keane, Elliott, & Sauseng, 2015; Gross et al., 2004).
36 This alternative alpha/beta entrainment hypothesis could explain the absence of the blink at in the 6.3 Hz and 36 Hz conditions. However, these results could be adequately explained by the two stage model of the blink by Chun and Potter (Chun & Potter, 1995). According to this model, T2 processing is delayed by the appearance of the first target, leading to subsequent interference when T2 +1 occurs ~100 ms later. In the theta (6.3 Hz) condition, both T1 and T2 accuracy were near ceiling, which could be attributed to the lack of effective backward masking from the T1+1 and T2+1 items (160 ms SOA). It has been well established that a sufficiently delayed T2+1 will lead to the absence of the blink, irrespective of the stream frequency. Likewise, the reduced or absent blink in the gamma (36 Hz) condition could due to the backward and/or forward masking of T2 and/or T1. In addition, the reduction in T1 accuracy in the gamma condition, combined with a three-alternative forced choice task (3AFC), likely led to the inclusion of a significant number of “T1 correct” trials in which T1 was not perceived (i.e., guessed correct T1), skewing the results of T2|T1. Finally, the magnitude of the blink produced at 10 and 16 Hz using the paradigm employed in Experiment 1a was small compared to most blink paradigms, putting to question the generalizability of the results.
Experiment 2
In Experiment 2, we sought to distinguish between the Alpha/Beta Entrainment account and the Chun and Potter account of the attentional blink, while addressing the additional concerns of the blink magnitude and generalizability of Experiment 1. First, we changed the task to the report of letters embedded in a series of task irrelevant digits, a paradigm known to produce a large
attentional blink. We also adjusted the timing of T2+1 to be equal in all conditions to more closely equate backward masking, removed the item immediately preceding T2 from the gamma condition to prevent excessive forward masking of T2, and adaptively adjusted the luminance of T1 to better match T1 performance between conditions. As the pre-T2 frequencies were largely maintained, as in
37 Experiment 1, the Alpha/Beta Entrainment account predicts a substantial blink only at 10 and 16 Hz, while the two-stage model predicts a blink for all conditions (See Figure 2.4).
Figure 2.4 Temporal trial design and predictions of Experiment 2 according to each account.
Methods
Experiment 2 was the same as Experiment 1 except as follows:
Twelve participants (mean age: 21.8 years, 9 female) participated in Experiment 3. Frequency order was counterbalanced across participants using a random Latin square design. The task was to report two letters presented among digits, in order to increase the depth of the attentional blink compared to Experiment 1. The stimuli were changed such that all distractors were randomly selected from the digits ‘1’ through ‘9’, with the constraint that the same digit would never be presented twice in a row within a trial. Targets were selected from all 26 letters of the English alphabet, with the
constraint that T1 and T2 would never have the same identity within a trial. Subjects were explicitly informed that ‘O’ should be viewed as a letter, not the number zero. Participants were told they
38 could report each target letter in any order, and responses matching either target were marked as correct at the position presented, regardless of the order in which the response was input.
Participants were forced to make two different letter responses before the trial would continue.
Pure chance T1 or T2 performance was 8%.
All digits and letters except for T1 were light grey (RGB: [128 128 128]) presented on a black
background. Courier font was used instead of Arial, due to poor spatial overlap between letters and numbers in Arial font (still ~1° visual angle). In order to better equate the overall accuracy of T2, the time between the onset of T1 and the following distractor was fixed at 97 ms for all conditions, and the distractor immediately preceding T2 (T2-1) in the Gamma condition was no longer presented (i.e.
the T2-1 SOA changed from 28 to 56 ms) to reduce the forward masking of T2. No other changes to the timing of the trial sequences were made (See Figure 2.4).
To approximately equate T1 performance across frequency, the luminance of T1 relative to all other items (relative contrast) was manipulated for each frequency to achieve 80% T1 accuracy at lag 700.
Pilot data was used to estimate this threshold and set initial relative contrast values to 34%, 57%, 146%, and 179% for the Theta, Alpha, Beta and Gamma conditions, respectively. Starting with these initial values, T1 contrast was adjusted in a 4-up, 1-down staircase procedure in increments of 20%, based solely on Lag 700 performance, but applied uniformly to all lag conditions. The maximum T1 contrast was capped at 200%. The final threshold values across the twelve experimental participants matched well with the initial settings (M: [34% 56% 137% 190%], SD: [13% 17% 32% 17%]). This contrast manipulation was effective at maintaining equal T1 Lag 700 accuracy in all frequency conditions excepting the Gamma condition (M: [82% 82% 82% 61%], SD: [4% 4% 5% 15%]). In the Gamma condition, nine of 12 participants had contrast thresholds at or above maximum allowed, resulting in reduced, though well above chance, T1 performance (chance = 8 %).
39 Each participant completed 78 trials per block, for a total of 312 trials participant. All 26 were letters pseudo randomly selected to occur exactly once at the T1 and T2 positions for each Lag, within each frequency block.
Analysis
A repeated measures one-way ANOVA across Frequency was performed for overall T1 accuracy and AB magnitude, along with all pairwise comparisons as well as tests for the presence of the blink at each frequency, as in Experiment 1.
Results
Overall T1 accuracy was significantly different between conditions (F(3,33) = 13.76, p = .001, partial η2 = .556), driven by relatively reduced accuracy in the gamma condition (p < .005 for all
comparisons between 36 Hz and all other frequencies). This resulted due to some participants failing to reach 80% performance, even at the maximum allowed contrast (200%) (See Figure 2.5).
Nevertheless, given that chance level performance in Experiment 2 was 8%, the T1 contrast manipulation in Experiment 2 achieved the goal of dramatically reducing the number of randomly guessed T1 correct responses.
40
Figure 2.5 Results of Experiment 2. T1|T2 accuracy across four frequencies and three SOAs. Error bars represent the across participant standard error.
A substantial blink was observed in all conditions (p < 0.005 at all frequencies, See Figures 2.5 and 2.6). The ANOVA testing for differences in AB magnitude between frequencies was marginally significant (F(3,33) = 2.59, p = .067, partial η2 = .166), with pairwise comparisons between
frequencies revealing a significantly lower AB magnitude at 6.3 Hz than 10 Hz (p = .048). No other pairwise comparisons approached significance.
41
Figure 2.6 Bar plot of AB magnitude across Frequency for Experiment 2. A substantial blink is observed at all frequencies.
Error bars represent the across participant standard error.
Discussion
Combined, the results of Experiments 1 and 2 are largely consistent with the two-stage model of the attentional blink, and are inconsistent with the Alpha/Beta Entrainment hypothesis of the
production of the attentional blink. In Experiment 2, a substantial blink can be observed, even when stimulating at theta (6.3 Hz) and gamma (36 Hz) frequencies. The absence of a significant blink at 6.3
production of the attentional blink. In Experiment 2, a substantial blink can be observed, even when stimulating at theta (6.3 Hz) and gamma (36 Hz) frequencies. The absence of a significant blink at 6.3